Accuracy and reliability of the RGB-D camera for measuring walking speed on a treadmill

RGB-D cameras (Red Green Blue+Depth) are widely employed in exergames designed to physically stimulate elderly people. Nevertheless, the intensity of the physical activity reached with the existing solutions is rarely sufficient to obtain a real impact on the physical fitness and thus on the health status of this population. In this context, a Point Cloud Based System (PCBS) has been developed to interface ordinary motorized treadmills with exergames through a simple RGB-D camera, to induce players to perform physical activities at higher intensities. The goal of this study was to assess the accuracy and reliability of PCBS to measure the walking speed of a subject on a standard motorized treadmill based on the image streams of an RGB-D camera.

36 participants performed three 10min walking exercises, divided in 5 blocks of 2min at the following constant ordered speeds: 0.42, 0.69, 0.97, 1.25 and 1.53ms(-1). The measured walking speeds are compared to those obtained through a Marker Based Control System (MBCS).

Results showed a high system accuracy (bias: 0.013±0.015ms(-1)), a good reliability (ICC=0.63-0.91) and a low variability (SEM=1-5%; MD=2.7-14%).

Accuracy and reliability of PCBS are consistent with those obtained in similar existing systems measuring gait parameters.

Within the context of the development of exergames, PCBS may be combined with exergames to perform physical activities at sufficiently high intensities in the elderly population, in order to improve their physical health and possibly prevent/delay cognitive impairment.